Witold Pedrycz

Department of Electrical & Computer Engineering

University of Alberta, Edmonton Canada

and
School of Computer Science, University of Nottingham, Nottingham,

NG7 2RD, UK.

and

Systems Research Institute, Polish Academy of Sciences

Warsaw, Poland

e-mail: pedrycz@ee.ualberta.ca

With the remarkably diversified plethora of design methodologies and algorithmic pursuits present today in system modeling and fuzzy modeling, we also witness a surprisingly high level of homogeneity in the sense that the resulting models are predominantly concerned with and built by using a data set coming from a single data source. 

 In this talk, we introduce a concept of collaborative granular modeling. In a nutshell, we are faced with a number of separate sources of data and the resulting individual models formed on their basis. An ultimate objective is to realize modeling at the global basis by invoking effective mechanisms of knowledge sharing and collaboration. In this way, each model is formed not only by relying on a data set that becomes locally available but also is exposed to some general modeling perspective by effectively communicating with other models and sharing and reconciling revealed local sources of knowledge.

 Several fundamental modes of collaboration (by varying with respect to the levels of interaction) are investigated along with the concepts of collaboration mechanisms leading to the effective way of knowledge sharing and reconciling or calibrating the individual modeling points of view. The predominant role of information granules with this regard is stressed.

 For illustrative purposes, the underlying architecture of granular models investigated in this talk is concerned with rule-based topologies <R_i, f_i> with R_i being a certain information granule (typically set or fuzzy set) formed in the input space and f_i denoting any local model realizing a certain mapping confined to the local region of the input space and specified by R_i.

 It is also shown that the collaboration and reconciliation of locally available knowledge ultimately give rise to the concept of higher type information granules, for instance fuzzy sets of type-2 and interval-valued fuzzy sets. With this regard, it is shown that the principle of justifiable granularity offers a constructive way of forming type-2 fuzzy sets.

Extracting usable and useful knowledge from large and complex data sets is a difficult and challenging problem. In this paper, we show how two complementary techniques have been used to tackle this problem in the context of breast cancer. Diagnosis concerns the identification of cancer within a patient; in contrast, prognosis concerns the prediction of the ongoing course of the disease, including issues such as the choice of potential treatments such as chemotherapy or drug therapy, in combination with estimation of chances (or length) of survival. Reliable prognosis depends on many factors, including the identification of the type of this heterogeneous disease. We first use a consensus clustering methodology to identify core, well-characterised sub-groups (or classes) of the disease based on a large database of protein biomarkers from over a thousand patients. We then use fuzzy rule induction and simplification algorithms to generate a simple, comprehensible set of rules for use in future model-based classification. The methods are described and their use is illustrated on real-world data.